Researchers offer a more complete, yet easily modified model for estimating American life expectancy by age and sex, incorporating for the first time the decline in tobacco use, increase in obesity and well-known trends and patterns of mortality.
Population mortality forecasts are an important underpinning for many analyses. Generally, attempts to incorporate more information into models have resulted in less accurate forecasts, so modelers have continued to use simple linear extrapolations based on limited data. But here, researchers propose an alternative Bayesian hierarchical forecasting model which allows for the inclusion of more data.
The authors observe their forecasts yield “venerable” demographic patterns, suggesting the model is statistically sound. Noting there is still much to learn about the demographic impact of risk factors, such as obesity and smoking, the authors believe the model provides a good base for future improvements because it is easy to modify.